Annual growth layers as proxies of past growth conditions for benthic microbial mats in a perennially ice-covered Antarctic lake


  • Editor: Riks Laanbroek

Correspondence: Donna L. Sutherland, National Institute of Water and Atmospheric Research, PO Box 8602, Christchurch, New Zealand. Tel.: +64 3 348 8987; fax: +64 3 348 5548; e-mail:


Perennial microbial mats can be the dominant autotrophic community in Antarctic lakes. Their seasonal growth results in clearly discernible annual growth layering. We examined features of live microbial mats from a range of depths in Lake Hoare, Antarctica, that are likely to be preserved in these layers to determine their potential as proxies of past growth performance. Cyanobacteria dominated the mat for all but the deepest depth sampled. Changes in areal concentrations of phycobilin pigments, organic matter and extracellular polysaccharide and in species composition did not correspond to changes in various water column properties, but showed a linear relationship with irradiance. Carbonate accumulation in the mats correlated with biomass markers and may be inferred as an index of mat performance. We examined the carbonate content of annual layers laid down from 1958–1959 to 1994–1995 in sediment cores from 12 m depth. The carbonate content in the layer showed a significant correlation with the mean summer air temperature. These data suggest a link between air temperature and microbial mat growth performance, and suggest that it is mediated via irradiance. Laminated microbial mats in Antarctic lakes have the potential to act as fine-resolution records of environmental conditions in the recent past, although interpretation is complex.


The McMurdo Dry Valleys of Antarctica are an extreme environment. They form one of the coldest and driest deserts on earth (Vincent, 2004); yet within this desert are a number of perennially ice-covered, endorheic, meromictic lakes, that are home to microbial-dominated communities that are uniquely acclimated to the extreme conditions that these lakes offer. While these lakes have a superficial appearance of persistence and stability, they are highly responsive to variation in regional climate on a range of time scales (Doran et al., 1994). In particular, the water balances of these closed-basin lakes are controlled by the balance between the influx of glacier melt during the austral summer and the loss from the lake surface via evaporation, sublimation and ablation. During cool periods with a low glacial melt, inflows decline while ablation persists and lake levels decrease and the lake-water solute concentrations increase until shallow brine pools may be all that remains. During periods of increased glacial melt, lake levels rise and solutes become more dilute, although in many instances inflowing meltwater has formed dilute caps on top of the dry-period brine pools (Spigel & Priscu, 1998). The net result has been cycles of filling and drying on millennial time scales as well as short-term (decadal) fluctuations in lake level and ice thickness of the order of meters (Chinn, 1993; Doran et al., 1994) and highly structured water columns.

Understanding how the benthic microbial communities that dominate the Dry Valley lakes respond to these changing physical conditions is a prerequisite to predicting their responses to future change and assessing the resistance and resilience of modern lake biota. While large-scale millennial changes must profoundly affect these communities, what is not well understood is how year-to-year and inter-decadal variability affect lake communities.

Linking past changes in environmental conditions to ecosystem performance requires either synchronous observational data or development of robust proxies for one or both. In this study, we explore the extent to which records laid down in the recent past in lacustrine sediments colonized by photosynthetic microbial mats can be related to growth conditions. We have shown how the perennial benthic microbial communities that are widespread in Dry Valley lakes (Wharton et al., 1983) form annual laminations when they grow undisturbed (Hawes et al., 2001), and it has been shown previously that these laminations persist deep into sediments (Squyres et al., 1991). Here we explore the possibility that the features of these annual growth layers may be used as proxies of past growth performance. We first examine the features of live mats as a ‘training set’ to understand links between growth and environment, and then examine three cores taken from a key transitional region of the lake to determine whether information in laminae could be related to climatic records covering the period during which the core was deposited.

Study site

Lake Hoare (77°38′S, 162°53′E) is a glacier-dammed, closed-basin lake with perennial ice cover situated in the Taylor Valley, Antarctica. The lake has a surface area of 1.94 km2, with maximum and mean depths of 34 and 14 m, respectively (Spigel & Priscu, 1998). The relatively thick ice cover (3–5 m) influences many aspects of the lake environment including the irradiance regime, gas exchange, and sediment deposition, and it eliminates wind-driven mixing, thus allowing a stratified water column to develop and persist (Wharton et al., 1989).

The irradiance regime is one of slowly changing but continuously low photon flux. In winter, there are 4 months of complete darkness. In summer, there are 4 months of continuous light but the ice transmits <1– 3.5% of this to the water column. The vertical extinction coefficient for downwelling irradiance within the water column below the ice to 22 m depth is typically 0.12–0.22 m−1 (Howard-Williams et al., 1998). The water column is density stabilized, with an inflection at 13–15 m depth, which divides the lake into upper and lower compartments. Compared with the lower compartment, the upper one is characterized by higher concentrations of dissolved oxygen (Wharton et al., 1986), lower concentrations of dissolved nutrients, particularly nitrate (Lizotte & Priscu, 1992), and higher pH (Cathey et al., 1981). Profiles of selected water column determinands for the period of this study were obtained from and are plotted as Fig. 1.

Figure 1.

 Selected determinands for Lake Hoare obtained from for November 2002, the main period of this study.

Microbial mats

Benthic microbial mats line the lake floor from the lake edge into the anoxic layer, at 25 m. The mats are comprised primarily of cyanobacteria, diatoms and bacteria, but differ in gross morphology and species composition with depth (Wharton et al., 1983). These mats are highly adjusted to low irradiance through efficient light harvesting and utilization, and compensation points of <1 μmol photons m−2 s−1allow net production to occur to >20 m depth in this lake (Hawes & Schwarz, 1999, 2000; Vopel & Hawes, 2006). Consumer organisms are limited to several protozoan taxa, rotifers, tardigrades and nematodes (Cathey et al., 1981; Parker et al., 1982). In the absence of bioturbation from macroscopic fauna, microbial mats are laminated on millimeter scales (Wharton et al., 1983). Model estimates of growth rates and experimental studies support the hypothesis that these horizontal laminations are annual growth layers (Hawes et al., 2001). Hawes & Schwarz (1999) recognized an ‘active layer’ comprising the upper three to nine laminae as containing all of the photosynthetically active components of the microbial mat, and this has been confirmed experimentally (Vopel & Hawes, 2006). Carbonate precipitates accumulate within the mats, notably as calcite crystals (Wharton et al., 1983).

Materials and methods

Characteristics of microbial mats along a depth gradient


Microbial mats were sampled in November–December 2002, at water depths of 8, 10, 12, 15, 18 and 22 m by a diver working under the ice cover. Divers cut through the active layer of the mats (sensu Hawes & Schwarz, 1999) and into the underlying material using a sharpened 10-cm plexiglass ring and carefully transferred the defined sample into opaque containers. These were sealed underwater, returned to the surface, immediately placed in a darkened, insulated box and taken to a darkened lakeside laboratory for further processing. In the laboratory, the cohesive active layer of each sample was separated from the underlying, unpigmented and weakly cohesive material and used in subsequent analyses. At lakeside, subsamples were examined by light microscopy and preserved in gluteraldehyde (EM grade) for more detailed identification of algae and cyanobacteria. Mat samples were also submerged in 5-cm-diameter dishes of water frozen intact for return with minimal structural disturbance to New Zealand for confocal laser scanning microscopy. Finally, five replicate samples (area 5.4 cm−2) were taken from the active layer at each depth, frozen and returned to New Zealand. On return to New Zealand, these samples were freeze dried and ground to a fine powder and weighed. Weighed aliquots of each sample were then taken for analysis of total carbonate and total organic matter, total carbohydrate, extracellular polymeric substances (EPS) and photosynthetic pigments.

Total carbonate and organic matter

Aliquots of freeze-dried microbial mat were weighed into 2-mL plastic centrifuge tubes and to each tube 1 mL of 10% HCl was added to displace carbonates. Acidified samples were then centrifuged, rinsed three times in distilled water to remove the resulting chlorides, redried and weighed. Weight loss was taken to indicate the carbonate content of the original sample. Samples were then carefully transferred to crucibles, combusted at 450 °C for 4 h, cooled in desiccators and the organic matter was estimated as loss of mass on ignition (LOI). Results were converted to mg cm−2 using the total sample and aliquot weights.

Total carbohydrate

Total carbohydrate was determined using the phenol–sulphuric acid assay (Dubois et al., 1956). Ultrapure water (2 mL) was added to a known weight of a freeze-dried microbial mat, followed by 1 mL of 5% aqueous phenol (w/v) and 5 mL of concentrated H2SO4. The absorbance was measured against a reagent blank at 485 nm and calibrated against glucose standard. Results were converted to glucose equivalents cm−2 using the total sample and aliquot weights.


Total EPS were extracted from weighed subsamples of a freeze-dried microbial mat in 100 mM Na2EDTA for 15 min at 20 °C and then precipitated overnight in cold 70% ethanol. The extract was then centrifuged, the supernatant was discarded and the resultant pellet was resuspended in ultrapure water. The carbohydrate content of the resuspension was estimated as glucose equivalents using the phenol–sulphuric acid method described above. Results were converted to glucose equivalents cm−2 using the total sample and aliquot weights.


Subsamples for pigment analysis were weighed and divided for chlorophyll a and phycobilin analyses. Chlorophyll a samples were extracted in 90% acetone at 4 °C, in dark, for 4 h. Samples were then centrifuged at 1250 g for 10 min and the absorbance of the supernatant was read on a Jasco 7850 double-beam spectrophotometer before and after acidification (Marker et al., 1980). Chlorophyll-a concentration was calculated after Marker et al. (1980) and expressed on a freeze-dried weight basis. Phycobilin pigments (phycocyanin and phycoerythrin) were extracted according to the methods described by Downes & Hall (1998). Briefly, samples were extracted in 0.05 M Tris buffer, in the dark, for 4 h and cleared by centrifugation. The fluorescence of the supernatant was measured on a Varian Cary Eclipse spectrofluorometer using excitation/emission wavelength combinations of 496/564 nm for phycoerythrin and 542/640 nm for phycocyanin. Fluorescence was calibrated using purified pigments (Sigma Chemicals). Results were converted to μg cm−2 using the total sample and aliquot weights.


Fresh subsamples of microbial mats for microscopy were drawn through a syringe to disperse the mat material and form a homogenous sample. Subsamples of the fresh material were viewed at × 1000 magnification, under oil immersion, on a Lecia DMLB microscope. Because of the nature of the masses of irregularly coiled fine trichomes (<1 μm wide), it was not possible to quantify individual species of cyanobacteria using direct counts. Instead, relative abundance was estimated as per cent cover along five random slide transects for each sample. Only diatom valves containing intact chloroplasts were considered in the relative abundance estimates. Identifications of the diatoms were primarily based on Krammer & Lange-Bertalot (1991, 1997a, b, 2000), and more specialized literature including Cremer et al. (2004), Kellogg & Kellogg (2002), Sabbe et al. (2003), Spaulding et al. (1997, 1999, 2008). Cyanobacteria taxonomy was based primarily on Komàrek & Anagnostidis (2005), and more specialized literature including Komàrek (2007) and Komàrek et al. (2008).

After examination of fresh material, further subsamples were prepared for more detailed diatom identification by oxidation in H2O2 and H2SO4 (Barber & Haworth, 1981). These were placed onto cover glasses, mounted in naphrax mounting medium and counted until 300 valves were enumerated and expressed as per cent diatom abundance.

Confocal laser scanning microscopy and image processing

Five replicate subsamples from each sampling depth of unstained, thawed intact microbial mat samples were viewed on a Leica TCS SP5 broadband Confocal Laser Scanning Microscope (CLSM) equipped with a × 20 (0.69 NA) water immersion objective. The microbial mats were illuminated with two lasers (DPSS 561 and HeNe 633) generating excitation wavelengths of 561 and 663 nm. These excitation wavelengths elicited fluorescence mostly from the phycobilin (561 nm) and chlorophyll (663 nm) pigments, making the organization of trichomes and diatoms clearly visible.

Subsamples of an active layer for confocal microscopy were viewed in cross section, and a series of fields of view along a transect were analysed to cover the full thickness of each section. For each field, 65 optical sections, each 0.32 μm thick, were recorded and a summa projection (all optical sections reconstructed to form one image) was performed of a cross-section 20.8 μm thick into the mat. The summa projections were then used to calculate the biovolume of cyanobacteria and diatoms within the active layer, converting fluorescing area to biovolume by assuming that diatoms and cyanobacteria were cylindrical. We note that most cells in these samples were full of pigment and that pigment area was a good proxy for cell area. Biovolume of the cross-section analysed was calculated using the computer-aided image analyses package image-j 1.38, with biovolume plugins (

Sediment cores

Sediment cores were collected during the austral summer of 1996–1997. Three cores were collected at 12 m depth using 6-cm-diameter plexiglass core tubes. The tubes were driven into the sediment, capped at both ends and returned to the surface, where they were frozen and returned to New Zealand for analysis. Cores were subsequently removed from the tubes and slowly thawed; it was found that these cores had retained their fine-scale-layered pattern with no signs of freeze disruption and distinct laminae were readily separable. The exceptions to this were the top two laminae (most recent growth), which were removed together and discarded. Individual laminae were freeze-dried, weighed and subsampled for total carbonate content and abundance of diatom species, using the methods described above. Mean carbonate contents (mg cm−2 per lamina)±SE were calculated for the three cores.

Statistical treatment of data

The quantitative data were examined using linear and nonlinear regression analysis using the statistical package sigmastat 3.5 (StatSoft Inc.). Where necessary, to remove asymmetric variance, variables were ln-transformed before analysis. Similarities were calculated using the Bray–Curtis similarity index, and nonmetric multidimensional scaling (NMDS) was used to construct a ‘map’ of similarity. Relationships between both cyanobacteria and diatom community composition and environmental variables were explored using the BIO-ENV procedure in the statistical package primer (Clarke & Warwick, 2001). BIO-ENV was considered to be suitable for exploring the present dataset because of the small number of samples from a restricted area (Clarke & Ainsworth, 1993). anosim was used to compare differences between diatom communities in the sediment core. To reduce the heteroscedasticity of the species counts, abundance data were transformed using ln (N+1). Multivariate analyses, including anosim, were carried out in primer v6 (Clarke & Warwick, 2001).


Composition of the active layer

The amount of organic matter, carbonate, total carbohydrate and EPS in the active layers of microbial mats varied with depth (Fig. 2). In general, a pattern was evident whereby determinands declined exponentially with increasing depth up to and including 18 m. Twenty-two meters was an outlier in all cases. Logarithmic regression analysis of depths from 8 to 18 m showed highly significant relationships, with exponents varying from −0.21 to −0.28 m−1, with ln(EPS) showing the weakest correlation with depth (Table 1). The proportionality of ln-transformed carbohydrate, EPS and carbonate to LOI tended to be constant, including the 22 m samples (Fig. 3). The possible exception to this is EPS, where an exponential proved a better fit than linear, indicating a proportionally higher EPS content at shallow depths.

Figure 2.

 Variation of bulk properties of the active layer of microbial mat from Lake Hoare with depth. All are plotted on log–linear axes.

Table 1.   Results of ln-linear regression analysis of various bulk properties of the active layer of microbial mat from Lake Hoare vs. depth
Property (units)NAdjusted r2Exponent (m−1)P of regression anova
  1. In all cases five replicate samples were analysed at each of five depths from 8–18 m and were treated as independent samples. Samples from 22 m depth have been excluded from this analysis.

Carbohydrate (glucose eq cm−2)250.887−0.24 ± 0.02<0.001
EPS (glucose eq cm−2)250.750−0.23 ± 0.03<0.001
Carbonate (mg cm−2)250.877−0.28 ± 0.02<0.001
Loss on ignition (mg cm−2)250.910−0.21 ± 0.01<0.001
Figure 3.

 Relationships between loss on ignition and ln-transformed total carbohydrate, carbonate and EPS. Filled symbols are from depths 8 to 18 m; open symbols are for 22 m. Solid lines are drawn by least-squares linear regression through all data points. Regression anovas were significant at P<0.0001, P<0.0018 and P<0.002 from left to right. In the right-hand panel, the dotted line is fitted by exponential regression and was a better fit than the linear, one with a regression anova of P=0.0003.

The relationship between collection depth and pigment content differed from that for the other determinands. Phycoerythrin was the most abundant pigment and this declined linearly with increasing depth, until it was virtually undetectable at 22 m depth (Fig. 4, Table 2). Phycocyanin also declined linearly with depth, although concentrations were low and the relationship was less robust than for phycoerythrin. Chlorophyll-a showed only a weak tendency to decline with depth, and if the 22 m samples were excluded there was no significant relationship.

Figure 4.

 Relationships between depth and various pigment contents measured of the active microbial layer from Lake Hoare.

Table 2.   Significance of linear regression relationships between depth and pigment concentrations in Lake Hoare, using the full depth range and excluding the 22 m samples
PigmentAll depthsExcluding 22 m
Adjusted r2NPAdjusted r2NP
  1. For all variables regression of ln-transformed variables yielded less significant relationships.


Species composition

The microbial mats from all depths were dominated by narrow, filamentous cyanobacteria (Table 3). Diatoms comprised only 10–15% of the community, with the exception of 22 m, where the relative proportions of diatoms and cyanobacteria switched (Figs 5 and 6). Leptolyngbya antarctica (West & West) Anagnostidis & Komárek was the dominant species within the mats at all depths. It comprised at least 70% of the photosynthetic community, increasing with depth to 85% at 18 m before declining to 15% at 22 m. Leptolyngbya fragilis (Gomont) Anagnostidis & Komárek., Pseudanabaena frigida (Fritsch) Anagnostidis and Leptolygbya angustissima (West & West) Anagnostidis & Komárek comprised the rest of the filamentous cyanobacterial community and were most frequent at the shallower depths.

Table 3.   Relative abundance of photosynthetic autotrophs present in the active layers of microbial mats along a depth profile in Lake Hoare
Species8 m10 m12 m15 m18 m22 m
Leptolygbya angustissima (West & West) Anagnostidis & Komárek (%)52<1<1
Leptolyngbya antarctica (West & West) Anagnostidis & Komárek (%)707075808515
Leptolyngbya fragilis (Gomont) Anagnostidis & Komárek (%)5555<1<1
Pseudanabaena frigida (Fritsch) (%)535<1<1
Diatoms (%)101515151585
Figure 5.

 Summa projection from confocal laser scanning microscopy of Lake Hoare benthic microbial mats at 12 m, showing the arrangement and density of trichomes mainly comprised of the cyanobacteria Leptolyngbya antarctica.

Figure 6.

 Summa projection from confocal laser scanning microscopy of Lake Hoare benthic microbial mats at 22 m, showing the dominant species Navicula contenta var. parallela.

The diatom community was more diverse than the cyanobacteria, with a species count of 16 species (Table 4). All depths supported between 13 and 15 species, with the exception of 22 m, where only nine species were recorded. The diatoms Psammothidium chlidanos (Horn & Hellerman) Lange-Bertalot, Navicula contenta var. parallela Petersen and Navicula gregaria Donkin were the most frequent species within the mats, together making up 59–84% of the counts at all depths (Table 4).

Table 4.   Mean abundance of diatoms in active layer of microbial mats from a range of depths in Lake Hoare
 8 m10 m12 m15 m18 m22 m
  1. Values are mean ± SE per cent of counts of 300 individuals from five replicate samples.

Diadesmis contenta (Grunow) DG. Round5.
Hantzschia abundans Lange-Bertalot1.
Hantzschia amphioxys (Ehrenberg) Grunow0.
Luticola gaussii (Heiden) DG. Mann3.
Luticola murrayi (West & West) DG. Mann4.
Luticola muticopsis fo. reducta (West & West) Spaulding2.
Microcostatus naumannii (Hustedt) Lange-Bertalot0.
Muelleria meridionalis Spaulding & Stoermer11.
Muelleria peraustralis (West & West) Spaulding & Stoermer1.
Navicula contenta var. parallela Petersen22.82.331.
Navicula gregaria Donkin8.92.615.81.439.
Navicula muticopsis fo. capitata Carlson3.
Navicula quaternaria Kellogg & Kellogg0.
Nitzschia westii Kellogg & Kellogg0.
Pinnularia cymatopleura West & West5.
Psammothidium chlidanos (Horn & Hellerman) Lange-Bertalot28.14.712.50.618.
Stauroneis cf. latistauros Ehrenberg2.

Two dimensional, nonmetric MDS ordination based on Bray–Curtis similarities of the diatom communities with depth showed that all diatom communities from 8 to 18 m were all at least 60% similar to each other, while 22 m was distinct from the other communities (Fig. 7). Within the 8–18 m group, the shallowest two sites (8 and 10 m) clustered as >80% similar to each other, as did the 12 and 15 m communities and these four depths clustered at the same point on the X-axis and formed a depth-wise sequence on the Y-axis. The 18 m community was identified as more distinct, lying to the right of the 8–15 m continuum, but was the least closely related to the 22 m community.

Figure 7.

 NMDS plots of ln-transformed diatom community composition along a depth gradient. ▴, 8 m; ▾, 10 m; ▪, 12 m; ♦, 15 m; •, 18 m; x, 22 m. The solid line represents 60% similarity and the dotted line represents 80% similarity.

Diatom taxa showing depth preferences, and thus responsible for the groupings seen in Fig. 5, were identified by anova. This showed that the relative abundance of N. contenta var. parallela was significantly higher at 22 m than at all other depths, high abundance of P. chlidanos at 18 m separated it from shallower communities and high abundance of N. gregaria separated 12 and 15 m from all other depths. Among other, less abundant taxa, Luticola spp. favoured shallow depths (8–10 m), while Pinnularia cymatopleura West & West favoured the deep mats (22 m). Other species did not show a preference for depth and occurred over the entire depth range, with the exception of 22 m, where a number of rare species were not found.

BIO-ENV analysis on all data produced a best rank correlation between the cyanobacteria and environmental matrices (temperature, pH, dissolved inorganic carbon (DIC), nitrate and free reactive phosphorus) of ρs=0.889 for nitrate. Addition of irradiance caused the match to deteriorate slightly (ρs=0.766), while other variables showed no correlation. This was mainly driven by the abrupt change in L. antarctica at 22 m. There were no correlations between diatom community composition and any of the environmental variables.

Photoautotrophic biovolume

The CLSM images were used to distinguish photosynthetic autotrophs containing fluorescing pigments from the rest of the mat matrix. Photoautotroph biovolume declined exponentially with depth (r2=0.51, P<0.001, Fig. 8). Biovolume at 8 m was lower than 10 m, although not significantly so, and the relationship between biovolume and depth improved when 8 m was excluded (r2=0.68, P<0.001). Abrupt changes in species composition at 22 m were confirmed with the CLSM analysis, where 89% of fluorescence was derived from diatoms (see Fig. 6).

Figure 8.

 Relationship between biovolume of live phototrophs within the benthic microbial mats, as determined by confocal laser scanning microscopy and depth.

Sediment core analysis

Carbonate content

A total of 36 layers were dissected out from the sediment cores. If these are annual layers, this represents a time series from the 1958–1959 Antarctic summer to 1994–1995. Carbonate content varied between layers, with a strong interannual variation. In order to better understand this apparent interannual variability, the data were compared with the mean daily summer air temperature records (November–February). The most extensive historic temperature records taken close to Lake Hoare are from nearby Scott Base, which coincidentally began in 1957–1958, while the record at Lake Hoare itself dates back only to 1987–1988 ( To determine whether Scott Base summer air temperatures are a useful proxy for temperature at Lake Hoare, we first compared these two datasets. There was a sufficiently good correlation (r2=0.654, P=0.015) to allow Scott Base temperature records to be used as a proxy. We compared carbonate cm−2 per layer with Scott Base summer air temperature anomaly, calculated as deviation from the long-term mean and with the Southern Ocean Oscillation Index (SOI) (obtained from the NOAA_CIRES Climate Diagnostics Center There was a strong positive correlation between carbonate content and the Scott Base air temperature anomaly (correlation coefficient=0.711, P<0.001, n=35) and a weaker but still significant one with SOI (correlation coefficient=0.403, P=0.016, n=35).

Diatom composition

Fourteen species of diatoms were identified in the core. All species, with the exception of Microcostatus naumanii (Hustedt) Lange-Bertalot, were also found in the 2002 microbial mat samples. Psammothidium chlidanos was dominant for most years, with N. contenta var. parallela, Muelleria meridionalis Spaulding & Stoermer and Hantzchia spp. subdominant on many occasions. Navicula gregaria, dominant in the modern mats from 12 m, was much less abundant in sediment layers. NMDS ordination based on Bray–Curtis similarities between the diatom communities from the core and from the modern mat samples showed that communities in all layers were only 60% similar to the 2002 assemblages from depths 8 to 18 m, while 22 m remained distinctly different from the rest (Fig. 9). anosim analyses showed that there was no significant correlation between diatom assemblages and carbonate or Scott Base temperature anomaly (r value≤0.25).

Figure 9.

 NMDS plots of ln-transformed diatom community composition within annual layers of sediment core. Diatom communities from live microbial mats along a depth gradient are plotted for comparison. • represents layers from sediment cores (estimated year labelled); ▾ represents live mats (depths labelled). The solid line represents 60% similarity, and the dotted line represents 80% similarity.


The under-ice water column of Lake Hoare contains a number of abrupt and less abrupt environmental gradients (Lyons et al., 1998; Fig. 1). Many of these gradients pivot around the depth range of 12–16 m where, with increasing depth, the temperature declines, nitrate concentrations increase and gradients of pH, DIC and nutrients other than nitrate are at their steepest. These vertical variations in water column properties are in turn set within a more continuous irradiance gradient. Benthic microbial mats occur across these gradients and understanding how mats are affected by these gradients allows an assessment of the extent to which their residues retained in lake sediments may be of use in hindcasting previous lake conditions.

In this study, we have looked at a mix of bulk properties of microbial mats, pigment contents and species composition. In general, we found that from 8 to 15 m, microbial communities showed no abrupt changes in species composition, but rather a gradual, depth-wise change, indicated in Fig. 7. At 18 m, and more spectacularly at 22 m, communities did change with a shift from cyanobacterial to diatom dominance, accompanied by a change in dominant diatom species to N. contenta var. parallela, at the expense of the cyanobacteria L. antarctica. Similarly, changes in bulk properties with depth were primarily gradual declines and it was only at 22 m that a deviation from a general exponential decline in loss on ignition, total carbohydrate, EPS, carbonate and active biovolume was seen. No discontinuous changes were seen that could be equated with the depths at which the most abrupt rates of change of temperature, nutrients, DIC or pH occurred.

The exponential decline in bulk ‘biomass’ properties of the mats suggests a relationship with energy supply more than with water quality. A near-exponential decline in irradiance is expected in any water column (Kirk, 1994), with the rate of exponential decline best described by the extinction coefficient K. The close similarity between the value of K for Lake Hoare of 0.2–0.3 (Howard-Williams et al., 1998) and the exponent of log-linear regressions of biomass proxies for the microbial mat in Table 1 (0.21–0.28) enhances circumstantial support for these declines being functions of energy supply. Irradiance is very low under the perennial ice cover of Lake Hoare (Hawes & Schwarz, 1999; Vopel & Hawes, 2006), and while communities have been shown to be very well acclimated to low irradiance, it is still likely to impose an overall limitation on growth rate. Phytoplankton in Dry Valley lakes responds to experimental nutrient enrichment in ways that suggest that nutrient limitation is a significant control on biomass dynamics (Priscu, 1995). While nutrients may also affect the growth of benthos, there is reason to believe that benthic communities are uncoupled from water column nutrient concentrations in these lakes. High concentrations of N and P have been recorded in interstitial mat waters from Lake Hoare (Quesada et al., 2008). Similarly, high interstitial nutrient concentrations are found in shallow-water Antarctic microbial mats (Villeneuve et al., 2001) and these also seem to be nutrient replete (Bonilla et al., 2005). Nutrient flux in ice-covered lakes is dominated by diffusion, which is an effective process over the short distances between decaying and growing material in the benthos, but much less so for planktonic communities.

Under a quasiequilibrium system, where active biomass loss is balanced by growth, rate limitation of biomass accrual can translate linearly to rate limitation of biomass. Such a system is rare in nature, but is likely to develop in microbial mats in Dry Valley lakes where communities are photosynthesizing under light-limiting conditions, at close to the maximum photosynthetic efficiency, most of the time and lack grazers or physical disruption (Hawes & Schwarz, 1999, 2000; Hawes et al., 2001; Vopel & Hawes, 2006). Where biomass-specific loss rate (LB– units per time) is a linear function of biomass (B– mass per area) and growth is independent of biomass but a linear function of limiting resource flux (R– resource units per area per time) and resource utilization efficiency (U– mass per resource unit), it is quickly apparent that biomass is linearly related to limiting resource supply, because, at ‘equilibrium’, biomass loss (BLB) equals biomass gain (RU)




and so where LB and U are constants, B is linearly related to R and where R is irradiance, this means B is exponentially related to depth. For this relationship to hold across a wide depth range, a similar community with similar values of parameters in Eqn. (2) must prevail. This appears to be the case for the upper 8–18 m of the lake, and it has been shown that photosynthetic parameters for communities from this range of depths are broadly similar (Hawes & Schwarz, 1999) and tend towards highly efficient absorption and utilization of irradiance. The break in this relationship at 22 m, coincident with changes in species composition, implies that the new community has markedly different parameters for Eqn. (2).

EPS provided the weakest correlation with depth of this group of variables and did not appear to suggest a relationship with declining irradiance as other biomass properties did. EPS secretion in diatoms and cyanobacteria can have a range of functions related to its various properties, including cell motility, ion exchange and binding, a diffusion barrier for toxins and heavy metals, attachment to surfaces and physical stabilization (Decho, 1990; Sutherland, 2001; Mancuso Nichols et al., 2005). Which of these is most important in the Lake Hoare microbial mats is unknown, although EPS-mediated motility is a feature of all taxa present, and a disproportionate decline of EPS relative to biomass with depth may reflect the changing role of EPS by the microbial mat or changes to community composition, including heterotrophic bacteria.

While it is reasonable that the ‘biomass’ indicators for the benthic communities (loss on ignition, carbohydrate) tended to decline exponentially, it is not immediately clear why precipitated carbonate followed this pattern, even through the pH and DIC concentration gradients between 12 and 15 m, which might be expected to change saturation kinetics. Dupraz et al. (2004) suggest that net microbial precipitation of calcium carbonate results from a temporal and spatial decoupling of the various microbial metabolic processes responsible for calcite formation and dissolution. Processes that favour precipitation include both oxygenic and anoxygenic photosynthesis, which locally elevate pH and hence shift the inorganic carbon equilibrium towards carbonate. In Antarctic systems, photosynthesis-driven pH increases can be expected to be seasonal and dependent on net photosynthesis, and so we might expect a relationship with irradiance in a light-limited regime. However, winter respiration might be expected to redissolve carbonates formed in this way, rather than leaving them to accumulate over time.

Binding of carbonate ions to EPS may in part be responsible for the retention of carbonate within the mat matrix and it is possible that EPS–calcium dynamics helps to integrate carbonate over time. The production and alteration of microbial EPS have been shown to be key processes in the precipitation and dissolution of calcium carbonate in microbial mats and biofilms as EPS-bound calcium is formed and released (Dupraz & Visscher, 2005).

The relatively high mat carbonate content at 22 m does not fit with a dependence on photosynthesis and suggests that other factors driving precipitation are disproportionately higher at this depth. The mats at 22 m are in close proximity to the anoxic layer at 25 m; it is likely that there is an increase in sulphur-reducing bacteria compared with other depths. While there are no known studies on the bacterial component of the microbial mats close to, or below the oxycline, Van Trappen et al. (2002) isolated 34 heterotrophic bacteria from microbial mats sampled from 3 to 4 m in Lake Hoare. The isolates included representatives from the genera Flavobacterium, Arthobacter and Pseudomonas, which all have sulphur-reducing species. The genus Flavobacterium appears primarily to play a role in remineralization processes and exhibits strong macromolecular hydrolytic capabilities (McCammon & Bowman, 2000). Sulphur-reducing bacteria have been shown to play a prominent role in carbonate precipitation and calcite formation through changes in the saturation index in the EPS matrix (Lyons et al., 1984; Visscher et al., 2000). An increase in sulphur-reducing bacteria would explain, in part, the disparity of the carbonate precipitate at 22 m relative to other depths, both with changes in the saturation index and increased EPS degradation. Calcium release, and hence potential calcite precipitation, follows the partial degradation of the EPS matrix (Arp et al., 2001), and the higher carbonate at 22 m may, indeed, be a measure of the increased heterotrophic degradation relative to other depths measured.

In contrast to biomass, pigment concentrations show quite different depth profiles, with a linear rather than an exponential decline. As has been found previously from this lake (Hawes & Schwarz, 1999), phycoerythrin is the dominant benthic pigment, well suited to absorption of the predominantly blue-green light that penetrates the ice cover (Hawes & Schwarz, 2000). This pigment shows a very strong linear decline with depth, quite unlike the near-depth independence of chlorophyll and the much weaker relationship of depth and phycocyanin. A linear decline of the dominant light-absorbing pigment may be expected in an irradiance-limited situation where a quasi-equilibrium biomass has been reached. If we assume that at any given depth the amount of pigment (P) deployed by a particular community type will be that which reduces irradiance to a constant very low but nonzero amount (Emin), this amounts to


where Ez is irradiance at depth z and SAC is the effective specific absorption coefficient for the pigment. Because Ez is equivalent to inline image, where Eo is irradiance immediately under the ice and Kd is the extinction coefficient for downwelling irradiance, therefore


indicating a tendency towards a linear depth/pigment concentration once an equilibrium biomass has been reached.

The cyanobacterial diversity was low and dominated by L. antarctica. This species is characteristic of frozen lakes in continental Antarctica and has also been recorded in lakes on James Ross Island (Komàrek & Elster, 2008). In spite of its wide distribution around Antarctica, little is known about the ecological tolerances of L. antarctica. Komàrek & Elster (2008) observed that L. antarctica mats developed a distinctly lower biomass in lakes on James Ross Island than that recorded in the continental lakes. There are little data from the James Ross Island study that are directly comparable to this study, but the total nitrogen concentrations in those lakes were up to eight times higher than 8–13 m in Lake Hoare, while only two times higher than 22 m in Lake Hoare. From our study, the most obvious explanation for the abrupt change at 22 m appears to be the light climate, but nutrients, such as nitrogen, may play a role to some degree. As L. antarctica dominates Antarctic lakes flora, it is important that the habitat range and ecological tolerances are understood.

All diatom species from the microbial mats have been previously recorded in Lake Hoare (Spaulding et al., 1997) and have been typically found in the Dry Valleys and other regions of Antarctica (Parker & Wharton, 1985; Howard-Williams et al., 1986; Vinocur & Pizarro, 2000; Kellogg & Kellogg, 2002; Sabbe et al., 2003; author's own data, unpublished). The total number of taxa was similar to that found in other Dry Valley lakes (author's own data, unpublished), but was low compared with lakes in other Antarctic and sub-Antarctic regions (Le Cohu & Maillard, 1986; Håkansson & Jones, 1994; Roberts & McMinn, 1999; Sabbe et al., 2003). Spaulding et al. (1997) found, in the upper 5 cm of core collected from 11 m in Lake Hoare, that N. gregaria, M. naumanii, Muelleria peraustralis (West & West) Spaulding & Stoermer and Pinnularia cymatopleura were greatest in abundance. Psammothidium chlidanos was greatest between 6 and 11 cm, while Diadesmis contenta (Grunow) DG. Round and N. contenta var. parallela were greatest in the lower part of the core.

In our study, while diatom species composition was similar in the sediment core and the 2002 active layer samples, the relative abundance was not. This was driven mainly by the absence of Hantzschia abundans in the 2002 samples and the significant reduction in abundance of N. gregaria in the sediment core. When these two species were removed from analysis, there was good agreement between the sediment core and the 2002 samples. In the 2002 samples, the diatom communities were related neither to the water column properties nor irradiance and our data provide no support for the use of diatoms in interpreting recent lake history. As the diatoms only constituted a small proportion of the microbial mat, it is likely that they are responding to the internal physio-chemical environment of the microbial mat and are less influenced by external factors such as the water column properties, or irradiance. Finer-scale measurements of the internal mat environment and diatom patch dynamics would be needed to confirm this.

Carbonate emerged as the best candidate as a proxy for microbial activity. It appeared to be positively correlated to microbial mat biomass variables and thus to irradiance. It is plausible to expect that the carbonate preserved in the core layers can be considered as fossils of metabolism. As discussed above, irradiance appears to be the overriding determining factor for biomass development in Lake Hoare, and fluctuations in the carbonate record may be an indication of changes in the light climate over time.

The strong correlation between the carbonate cm−2 per layer with Scott Base summer air temperature anomaly suggests that there may be a correlation between summer air temperatures and irradiance, although the exact relationship and mechanism are unclear. The amount of irradiance that is able to reach the microbial mats is influenced by cloud cover, ice thickness and optical properties, seasonal changes in water clarity and snow cover. How each of these interact and influence irradiance at depth will vary from year to year and all can be expected to be related to local climate. A study on the ice cover thickness of Lake Hoare over a 10-year period (1978–1988) found that it was progressively thinning at a rate exceeding 20 cm year−1 (Wharton et al., 1992), and a relationship between the peak summer temperature and the volume of glacier-derived meltwater entering the lake was considered a possible explanation for the thinning of the lake ice. Wharton et al. (1992) suggested that the extent of summer melting, consistent with summer temperature, was most likely responsible for the change in the ice cover thickness in Lake Hoare. However, while thinning ice may increase under-ice irradiance, warm temperatures promote meltwater flow that may decrease water clarity and increase lake level, which may negate the increased transmission of light through the thinning ice. Optical properties of the ice cover, including snow accumulation, ice crystal structure, the amount of sediment and air spaces within the ice, also vary with ice temperature (Howard-Williams et al., 1998), and all these factors are likely to also differ between warm and cool periods. Bertler et al. (2006) found that snow accumulations in the Dry Valleys were significantly correlated to the mean Scott Base summer temperatures, with high snow accumulations occurring during warmer, wetter periods, although snow cover on the lake surface typically does not persist throughout the season.

In this study, we have shown a correlation between the biomass variables and carbonate precipitate in the live mats. This suggests that the carbonate precipitate found in the annual growth layers in sediment cores serves as a good proxy for past growth performance. Based on comparisons between live mats and the physio-chemical properties of the lake, this growth performance would most likely reflect past irradiance levels within the lake rather than other water column properties, such as pH, nitrate, temperature, and DIC. Correlations between summer air temperature and carbonate accumulation suggest that the latter may indeed be a useful proxy for growth conditions, although further studies are needed to determine exactly how the local climate influences growth conditions, particularly irradiance, within the lake. The relationships between carbonate precipitate in other microbial mats and irradiance need to be spatially resolved both within lakes of varying light attenuation within the McMurdo Dry Valleys and in other regions throughout Antarctica, where the importance of different external drivers may vary.


This study was supported by the New Zealand Foundation for Research, Science and Technology (CO1605), Antarctica New Zealand, and the McMurdo Dry Valley Long-Term Ecological Research Program of the US National Science Foundation (OPP-9814972) funded and provided facilities and support for the field research. We thank Marie Uhle for assistance in the field and Kathy Walter for compiling Scott Base temperatures. Clive Howard-Williams and two anonymous reviewers provided useful comments on earlier drafts.